Lambda provides GPU cloud instances tailored for deep learning workloads. You can launch instances with NVIDIA A100, H100, and other high-end GPUs, pre-installed with CUDA, PyTorch, and TensorFlow. The pricing is straightforward — a fixed hourly rate per GPU type, with no egress fees on some plans. They also sell pre-configured GPU workstations for on-premises use.

It's used by AI teams at companies and research labs that need reliable GPU capacity without the complexity of general-purpose cloud platforms. The setup is minimal: choose a GPU, choose an image, launch. For training large models or running inference at scale, it's one of the more focused options in the GPU cloud market.